For the Wiley Encyclopedia of Electrical and Electronics Engineering

Underwater Acoustic Communication

Milica Stojanovic

Department of Electrical and Computer Engineering

Northeastern University

Boston� MA �����

Keywords� acoustic� communications� coherent� equalization� channel estimation� multi�

path� diversity combining� beamforming� multiuser detection� underwater data networks�

sparse equalization� blind equalization� multirate adaptive ltering� cyclostationary statis�

tics

The need for underwater wireless communications exists in applications such as remote

control in o��shore oil industry� pollution monitoring in environmental systems� collection of

scientic data recorded at ocean�bottom stations� speech transmission between divers� and

mapping of the ocean �oor for detection of objects� as well as for the discovery of new re�

sources Wireless underwater communications can be established by transmission of acoustic

waves Underwater acoustic communications are a rapidly growing growing eld of research

and engineering as the applications� which once were exclusively military� are extending into

commercial elds The possibility to maintain signal transmission� but eliminate physical

connection of tethers� enables gathering of data from submerged instruments without human

intervention� and unobstructed operation of unmanned or autonomous underwater vehicles

UUVs� AUVs�

Acoustic waves are not the only means for wireless transmission of signals under water

However� radio waves that will propagate any distance through conductive sea water are the

extra low frequency ones �� Hz���� Hz� which require large antennae and high transmitter

powers Optical waves do not su�er so much from attenuation� but they are a�ected by

scattering Consequently� transmission of optical signals requires high precision in pointing

the narrow laser beams While the laser technology is still being perfected for practical use�

acoustic waves remain the single best solution for communicating under water in applications

where tethering is unacceptable

The idea of sending and receiving information under water� is traced back all the way to

the time of Leonardo Da Vinci� who is quoted for discovering the possibility to detect a distant

ship by listening on a long tube submerged under the sea In the modern sense of the word�

underwater communications began to develop during the second World War� for military

purposes One of the rst underwater communication systems was an underwater telephone�

which was developed in ���� in the United States for communicating with submarines ���

This device used a single�sideband SSB� suppressed carrier amplitude modulation in the

frequency range of ���� kHz� and it was capable of sending acoustic signals over distances

of several kilometers However� it wasn�t until the development of VLSI technology that a

new generation of underwater acoustic communication systems began to emerge With the

availability of compact DSPs with their moderate power requirements� it began possible for

�

the rst time to implement complex signal processing and data compression algorithms at

the submerged ends of an underwater communication link

During the past few years� signicant advencements have been made in the development

of underwater acoustic communication systems ���� in terms of their operational range and

the data throughput Acoustically controlled robots have been used to replace divers in

performing maintenance of submerged platforms ����� high�quality video transmission from

the bottom of deepest ocean trenches ���� km� to the surface ship was established ����� and

data telemetry over horizontal distances in excess of ��� kilometers was demonstrated ����

As e�cient communication systems are developing� the scope of their applications contin�

ues to grow� and so do the requirements on the system throughput and performance Many

of the developing applications� both commercial and military� are now calling for real�time

communication with submarines and autonomous underwater vehicles� not only in point�to�

point links� but also in network congurations The emerging communication scenario in

which the modern underwater acoustic systems will operate is that of an underwater data

network consisting of both stationary and mobile nodes This network is envisaged to pro�

vide exchange of data� such as control� telemetry and eventually video signals� between many

network nodes The network nodes� both stationary and mobile ones� located on underwater

vehicles and robots� will be equipped with various sensors� sonars and video cameras A

remote user will be able to access the network via a radio link to a central node based on a

surface station

Towards achieving these goals� current research is focusing on the development of e��

cient communications and signal processing algorithms� design of e�cient modulation and

coding schemes� multiple access methods� and techniques for mobile underwater communi�

cations In the related eld of communication networking� the design of protocols� suited

for long propagation delays and strict power requirements encountered in the underwater

environment is being addressed Finally� in the eld of image processing� the development of

data compression algorithms suitable for low�contrast underwater images will enable image

transmission through band�limited underwater acoustic channels ���

�

��� System requirements

The achievable data throughputs� and the reliability of the underwater acoustic communi�

cation system� as measured by the bit�error probability� vary from system to system� but

are always subject to bandwidth limitations of the ocean channel Unlike in the majority of

other communication media� the use of underwater acoustic resources has not been regulated

yet by standards

In the existing systems� there are usually four kinds of signals that are transmitted�

control� telemetry� speech and video signals

Control signals include navigation� status information� and various on�o� commands for

underwater robots� vehicles and submerged instrumentation such as pipeline valves or deep

ocean moorings The data rates up to about � kilobit per second kbps� are su�cient for

these operations� but high reliability is required

Telemetry data is collected by submerged acoustic instruments such as hydrophones�

seismometers� sonars� or various other instruments which collect information about currents�

tides� pollution� etc� and it also may include very low rate image data Data rates on the

order of one to several tens of kbps are required for these applications The reliability

requirements are not so stringent as for the command signals� and a probability of bit error

of ���� � ���� is acceptable for many of the applications

Speech signals are transmitted between divers and a surface station or among divers

While the existing� commercially available diver communication systems mostly use analog

communications� based on single�sideband modulation of the � kHz audio signal� research is

advancing in the area of synthetic speech transmission for divers� as digital transmission is

expected to provide better reliability Transmission of digitized speech by linear predictive

coding LPC� methods requires rates on the order of several kbps to achieve close�to�toll

quality The bit error rate tolerance of about ���� makes it a viable technology for poor

quality band�limited underwater channels ���� ���

Video transmission over underwater acoustic channels requires extremely high compres�

sion ratios if an acceptable frame transmission rate is to be achieved Fortunately� underwater

images exhibit low contrast and detail� and preserve satisfactory quality if compressed even

�

to � bits per pixel Compression methods� such as the JPEG Joint Photographic Experts

Group� standard discrete cosine transform� have been used to transmit ��� � ��� pixel

still images with � bits per pixel� at transmission rates of about one frame per �� second

���� Further reduction of the required transmission rate seems to be possible by using ded�

icated compression algorithms� eg� the discrete wavelet transform ���� On the other hand�

underwater acoustic transmission of television�quality monochrome video would require com�

pression ratios in excess of ������ Hence� the bit rates required for video transmission vary

between �� kbps and ��� kbps Performance requirements are moderate� as images will have

satisfactory quality at bit error rates on the order of ���� � ����

� Channel characteristics

Sound propagation under water is primarily determined by transmission loss� noise� rever�

beration� and temporal and spatial variability of the channel Transmission loss and noise

are the principal factors determining the available bandwidth� range and signal�to�noise ra�

tio Time�varying multipath in�uences signal design and processing� often imposing severe

limitations on the system performance

��� Range� bandwidth and SNR

Transmission loss is caused by energy spreading and sound absorption While the energy

spreading loss depends only on the propagation distance� the absorption loss increases not

only with range but also with frequency� thus setting the limit on the available bandwidth

���

In addition to the nominal transmission loss� link condition is largely in�uenced by the

spatial variability of the underwater acoustic channel Spatial variability is a consequence

of the channel behavior as a waveguide� which results in various phenomena� including for�

mation of the shadow zones Transmission loss at a particular location can be predicted

by many of the propagation modeling techniques ��� with various degrees of accuracy Spa�

tial dependence of transmission loss imposes particularly severe problems for communication

with moving sources or receivers

�

Noise observed in the ocean exhibits strong frequency�dependence as well as site�dependence

Generally� the inshore environments� such as marine work�sites� are much noisier than deep

ocean� due to the man�made noise Unlike the man�made noise� most of the ambient noise

sources can be described as having a continuous spectrum and Gaussian statistics ��� As a

rst approximation� the ambient noise power spectral density is commonly assumed to decay

at �� dB�decade� both in shallow and deep water� over frequencies which are of interest to

communication systems design

Ambient noise� together with frequency�dependent transmission loss� determines the re�

lationship between the available range� bandwidth and SNR at the receiver input This

dependence is illustrated in Fig� which shows the frequency dependent portion of SNR for

several transmission ranges The SNR is evaluated assuming spherical spreading� absorp�

tion according to Thorp ��� and a �� dB�dec decay of the noise power spectral density�

Evidently� this dependence in�uences the choice of a carrier frequency for the desired trans�

mission range From the gure� the relationship between the available range and frequency

band also becomes apparent Underwater acoustic communication links can be classied

according to range as very long� long� medium� short and very short links For a long�range

system� operating over ������ km� the bandwidth is limited to few kHz for a very long dis�

tance on the order of ���� km� the available bandwidth falls below a kHz� A medium�range

system operating over ���� km has a bandwidth on the order of �� kHz� while only at very

short ranges below about ��� m� more than a hundred kHz of bandwidth may be available

Within this limited bandwidth� the signal is subject to multipath propagation through

a channel whose characteristics vary with time and are highly dependent on the location of

the transmitter and receiver The multipath structure depends on the link conguration�

which is primarily designated as vertical or horizontal While vertical channels exhibit little

multipath� horizontal channels may have extremely long multipath spreads Most notable

in the long� and medium�range channels� multipath propagation causes severe degradation

of the acoustic communication signals Combating the underwater multipath to achieve a

high data throughput is without exception considered to be the most challenging task of an

underwater acoustic communication system

�

��� Multipath

In a digital communication system which uses a single carrier� multipath propagation causes

intersymbol interference ISI�� and an important gure of merit is multipath spread in terms

of symbol intervals While typical multipath spreads in the commonly used radio channels

are on the order of several symbol intervals� in the horizontal underwater acoustic channels

they increase to several tens� or a hundred of symbol intervals for moderate to high data

rates For example� a commonly encountered multipath spread of �� ms in a medium�range

shallow water channel� causes the ISI to extend over ��� symbols if the system is operating

at a rate of �� kilosymbols per second ksps�

The mechanisms of multipath formation in the ocean are di�erent in deep and shallow

water� and also depend on the frequency and range of transmission Understanding of

these mechanisms is based on the theory and models of sound propagation Depending on

the system location� there are several typical ways of multipath propagation It is mostly

the water depth that determines the type of propagation The denition of shallow and

deep water is not a strict one� but usually implies the region of continental shelves� with

depth less than about ��� m� and the region past the continental shelves� respectively Two

fundamental mechanisms of multipath formation are re�ection at the boundaries bottom�

surface and any objects in the water�� and ray bending sound speed is a function of depth�

and the rays od sound always bend towards regions of lower propagation speed� If the water

is shallow� such as in the littoral region and the region of continental shelves� propagation

will occur in surface�bottom bounces in addition to a possible direct path If the water is

deep� as in the regions past the continental shelves� the sound channel may form by bending

of the rays toward the location where the sound speed reaches its minimum� called the axis

of the deep sound channel Since there is no loss due to re�ections� sound can travel in

this way over several thousands of kilometers Alternatively� the rays bending upwards may

reach the surface focusing in one point where they are re�ected� and the process is repeated

periodically The region between two focusing points on the surface is called a convergence

zone� and its typical length is �� ���� km

The geometry of multipath propagation and its spatial dependence are important for

�

communication systems which use array processing to suppress multipath eg� ����� �����

The design of such systems is accompanied by the use of a propagation model for predicting

the multipath conguration

Ray theory and the theory of normal modes provide basis for such propagation modeling

Recent references commonly use ray tracing for determining the coarse multipath structure

for communication channel modeling ��� �� A di�erent class of underwater acoustic com�

munication systems has recently been developed ����� which do not rely on the particular

multipath geometry� and is equally applicable in a variety of channels� regardless of the pa�

rameters such as range�to�depth ratio which determine the angles of incidence of multipath

arrivals

��� Time�variation

Associated with each of the deterministic propagation paths macro�multipaths�� which can

be modeled accurately� are random signal �uctuations micro�multipath�� which account

for the time�variability of the channel response Some of the random �uctuations can be

modeled statistically ������� These include surface scattering due to waves� which is the

most important contributor to the overall time variability of the shallow water channel In

deep water� in addition to surface scattering� internal waves contribute to the time�variation

of the signal propagating along each of deterministic paths

Surface scattering is caused by the roughness of the ocean surface If the ocean were calm�

a signal incident on the surface would be re�ected almost perfectly� with only deformation

being the phase shifting by � However� waves act as the displacement of the re�ection

point� resulting in the dispersion of energy Vertical displacement of the surface can be

well modeled as a zero�mean Gaussian random variable� with power spectrum depending

only on the wind speed ��� Motion of the re�ection point results in frequency spreading of

the surface�re�ected signal� signicantly larger than that caused by many other phenomena

Doppler spread of a signal component of frequency f caused by a single surface�re�ection

occurring at an incidence angle � is �������c�fw���cos� where c is the speed of sound�

nominally taken to be ���� m�s� and w is the wind speed in m�s ��� This dependence is

illustrated in Fig�� in which it is assumed that cos� � �� as it will be the case for a great

�

majority of communication links in which the range is much greater than depth Wind speed

of �� m�s corresponds to moderate winds Doppler spreads as high as several tens of Hz

are most likely to be found in short range links� which use relatively high frequencies For

longer ranges� at which lower frequencies are used� the Doppler spread will be lower� however�

multipath spread will increase Nevertheless� the channel spread factor� ie the product of

the Doppler spread and the multipath spread can be expected to decrease with range

As an example� Figs��� each show an ensemble of channel impulse responses� observed as

functions of delay over an interval of time These responses are estimated from experimental

measurements Relevant system parameters are indicated in the gures

While there exists a vast knowledge of both deterministic and statistical modeling of

sound propagation underwater� the implications this knowledge bears on the communication

channel modeling has only recently received more attention eg� ��������� A time�varying

multipath communication channel is commonly modeled as a tapped delay line� with tap

spacing equal to the reciprocal of twice the channel bandwidth� and the tap gains modeled

as stochastic processes with certain distributions and power spectral densities While it is

known that many radio channels t well within the model of Rayleigh fading� where the

tap gains are derived from complex Gaussian processes� there is no single model accepted to

date for any of the underwater acoustic channels Among the underwater acoustic channels�

modeling of the shallow water medium�range channel has received most attention� as this

channels is known to be among the most rapidly varying ones Most authors consider that

this channel is fully saturated� meaning that it exhibits Rayleigh fading ���� ���� ��� The deep

water channel has also been modeled as a Rayleigh fading channel� however� the available

measurements are scarce The question of statistical channel modeling is still a controversial

one� as some authors nd that both the deep and the shallow water channel t better within

a deterministic model ���

The factor that determines the performance of a digital communication system on a

frequency�spread channel is the Doppler spread normalized by the symbol rate In under�

water acoustic channels� the normalized Doppler spread can approach the limiting value of

���� for reliable coherent channel tracking And indeed� many researchers still nd that due

to surface variability a reliable carrier phase reference cannot be obtained in the shallow

�

water channels ����� ���� However� the implications time�varying multipath bears on the

communication system design are twofold On one hand� signaling at a high rate short

pulses� causes many adjacent symbols to interfere at the receiver� and requires sophisticated

processing to compensate for the ISI On the other hand� as pulse duration becomes shorter�

channel variation over a single symbol interval becomes negligible This allows an adaptive

receiver to e�ciently track the channel on a symbol�to�symbol basis� provided� of course� a

method for dealing with the resulting time�dispersion Hence� the time�varying multipath

causes a trade�o� in the choice of signaling rate for a given channel Experimental results

obtained on a shallow water medium�range channel ���� demonstrate these observations

The statistical channel measurements available today focus mostly on stationary com�

munication scenarios In a mobile underwater acoustic channel� vehicle speed will be the

primary factor determining the time�coherence properties of the channel� and consequently

the system design Knowledge of a statistical channel model has proven to be useful in

the design and analysis of land�mobile radio systems� however� it remains for the future to

develop such models for underwater mobile acoustic channels

� System design

To overcome the di�culties of time�varying multipath dispersion� the design of commercially

available underwater acoustic communication systems has relied so far mostly on the use of

noncoherent modulation techniques and signaling methods which provide relatively low data

throughput Recently� phase�coherent modulation techniques� together with array processing

for exploitation of spatial multipath diversity� have been shown to provide a feasible means

for a more e�cient use of the underwater acoustic channel bandwidth These advancements

are expected to result in a new generation of underwater communication systems� with at

least an order of magnitude increase in data throughput

Approaches to system design vary according to the technique used for overcoming the

e�ects of multipath and phase variations Specically� these techniques may be classied

according to �� the signal design� ie the choice of modulation�detection method� and

�� the transmitter�receiver structure� ie the choice of array processing method and the

�

equalization method� if any In this section� the design of several systems which have been

implemented is described While most of the existing systems operate on the vertical� or

the very short�range channels� the systems under development often focus on the severely

spread horizontal shallow water channels Signal processing methods used in these systems

are addressed in the following section

��� Systems based on noncoherent modulation

Noncoherent detection of FSK frequency shift keying� signals has been used for channels

exhibiting rapid phase variation such as the shallow water long� and medium�range channels

To overcome the ISI� the existing noncoherent systems employ signal design with guard times�

which are inserted between successive pulses to ensure that all the reverberation will vanish

before each subsequent pulse is to be received The insertion of idle periods of time obviously

results in a reduction of the available data throughput In addition� due to the fact that

fading is correlated among frequencies separated by less than the coherence bandwidth the

inverse of the multipath spread�� it is desired that only those frequency channels� separated

by more than this amount be used at the same time This requirement further reduces the

system e�ciency unless some form of coding is employed so that the adjacent� simultaneously

transmitted frequencies belong to di�erent codewords A representative system ���� for

telemetry at a maximum of � kbps uses a multiple FSK modulation technique in the ��

kHz��� kHz band This band is divided into �� subbands� in each of which a ��FSK signal

is transmitted Hence� out of a total of �� channels� �� are used simultaneously for parallel

transmission of �� information bits � information bits per one ��channel subband� This

system has successfully been used for telemetry over a � km shallow water horizontal path�

and a � km deep ocean vertical path It was also used on a ��� m shallow water path�

where probabilities of bit error on the order of ���� � ���� were achieved without coding

The system performance may be improved by using error correction coding� however� its

data rate will be reduced The multiple FSK system has been implemented ����� and is

commercially available with a maximum data rate of ���� bps Despite the fact that these

systems have bandwidth e�ciency which does not exceed �� bps�Hz� noncoherent FSK

is a good solution for applications where moderate data rates and robust performance are

��

required As such� these methods are being further developed� and a system has recently been

implemented ���� which uses orthogonal frequency division multiplexing OFDM� realized

with DFT�based lter banks This system was used on a medium�range channel� however�

due to the high frequency separation among the channels only every fourth channel is used�

and relatively long guard times �� ms guard following a �� ms pulse�� needed to compensate

for the multipath fading distortion� the e�ective data rate is only ��� bps

��� Systems based on coherent modulation

With the goal of increasing the bandwidth e�ciency of an underwater acoustic communica�

tion system� research focus over the past years has shifted towards phase�coherent modula�

tion techniques� such as PSK phase shift keying� and QAM quadrature amplitude mod�

ulation� Phase�coherent communication methods� previously not considered feasible� were

demonstrated to be a viable way of achieving high�speed data transmission over many of

the underwater channels� including the severely time�spread horizontal shallow water chan�

nels ��������� The new generation of underwater acoustic communication systems� based

on the principles of phase�coherent detection techniques� is capable of achieving raw data

throughputs that are an order of magnitude higher than those of the existing noncoherent

systems

Depending on the method for carrier synchronization� phase�coherent systems fall into

two categories� di�erentially coherent and purely phase�coherent The advantage of using

di�erentially encoded PSK DPSK� with di�erentially coherent detection is the simple carrier

recovery it allows Its disadvantage is performance loss as compared to coherent detection

While bandwidth�e�cient methods have successfully been tested on a variety of channels�

real�time systems have mostly been implemented for application in vertical and the very

short range channels� where little multipath is observed and the phase stability is good

In the very short range channel� where bandwidth in excess of ��� kHz is available� a

representative system ���� operates over �� m at a carrier frequency of � MHz and a data rate

of ��� kbps This system is used for communication with an undersea robot which performs

maintenance of a submerged platform A ���QAM format is used� and the performance

is aided by an adaptive equalizer A linear equalizer LE�� operating under an least mean

��

squares LMS� algorithm su�ces to reduce the bit error rate from ���� to ���� on this

channel

The current state�of�the�art in coherently modulated systems which use the deep ocean

vertical path is represented by an image transmission system ���� developed in Japan It

is a di�erentially coherent� ��DPSK system with carrier frequency of �� kHz� capable of

achieving �� kbps bottom to surface transmission over ���� m The eld tests of this system

indicate the achievable bit error rates on the order of ���� with linear equalizer operating

under an LMS algorithm

Another example of a successfully implemented system for vertical path transmission is

that of an underwater image and data transmission system developed in France ���� This

system uses a binary DPSK modulation at a data rate of ��� kbps The carrier frequency

of �� kHz was used for transmission over ���� m

Recent advances in digital underwater speech transmission are represented by a prototype

system described in ���� This system uses a code excited linear prediction CELP� method

to transmit the speech signal at � kbps The modulation method used is quadrature DPSK

A decision�feedback equalizer� operating under LMS algorithm is being used in the pool

tests Field tests have not been reported yet A similar approach is considered in ����

For the applications in shallow water medium�range channel� a DPSK system ���� uses

a direct�sequence spread spectrum method to resolve a strong surface re�ection observed in

the � km long� �� m deep channel The interfering re�ection is only rejected� and not used

for multipath recombining Data throughput of ��� bps in a spreading bandwidth of �� kHz

is achieved Such high spreading ratios are justied in interference�suppression applications

Current state�of�the art in phase�coherent communication systems is represented by the

prototype system implemented at the Woods Hole Oceanographic Institution in the USA

���� This system is based on purely phase�coherent modulation and detection principles

given in ���� The modulation format is QPSK� and the signals are transmitted at � kbps�

using a carrier frequency of �� kHz The system�s real�time operation in conguration as

a six�node network was demonstrated in the under�ice shallow water environment The ISI

caused by the shallow water multipath is processed by a decision�feedback equalizer operating

under an RLS recursive least squares� algorithm

��

� Signal processing methods for multipath compen�

sation

To achieve higher data rates� the more sophisticated systems based on phase�coherent sig�

naling methods must allow for considerable ISI in the received signal These systems employ

either some form of array processing� or equalization methods� or a combination thereof� to

compensate for the ISI Three main approaches have recently been taken towards this end

The rst two approaches use di�erentially coherent detection and rely on array processing

to eliminate� or reduce multipath The third approach is based on purely phase�coherent

detection and the use of equalization together with array processing for exploitation of the

multipath and spatial diversity

Array processing for multipath suppression has been used both at the transmitter and

at the receiver end An approach pursued at the University of Birmingham� UK ����

focuses on the use of transmitter arrays to excite only a single path of propagation Long

arrays and careful positioning are required to ensure complete absence of multipath� since no

equalization is used in this system Equalization is not deemed feasible in this work� since it

is considered that the shallow water test channel exhibits Rayleigh fading which is too rapid

for an adaptive equalizer to follow Instead� a receiving array is employed to compensate for

the possible pointing errors Binary and quaternary DPSK signals were used achieving data

rates of �� kbps and �� kbps� respectively The estimated bit error rate BER� was on the

order ���������� depending on the actual channel length In general� it was found that the

technique proposed is more e�ective at shorter ranges

An approach for multipath rejection using adaptive beamforming at the receiver end is

being investigated at the University of Newcastle� UK ����� The prototype beamformer

uses an LMS type of algorithm to adaptively steer nulls in the direction of a surface re�ected

wave Similarly as in the case of the transmitter array� it was found that the beamformer

encounters di�culties as the range increases relative to depth To compensate for this e�ect�

the use of an equalizer was proposed to complement the performance of the beamformer ����

The equalizer is of a decision�feedback type� and it operates under the LMS algorithm whose

low computational complexity permits real�time adaptation at the symbol rate A separate

��

waveform is transmitted at twice the data rate for purposes of time�synchronization The

system was tested in shallow water at �� kbps� and showed the estimated BER of ����

without� and ���� with the equalizer

The nal approach we shall describe is based on the use of purely phase�coherent de�

tection methods� which have been developed at Northeastern University and the Woods

Hole Oceanographic Institution� Massachusetts� USA ��������� for application in underwa�

ter acoustic channels with severe multipath These signal processing methods are based on

joint synchronization and equalization for combating the e�ect of phase variations and ISI

The method of fractionally spaced decision�feedback equalization is used with an RLS algo�

rithm It also incorporates spatial signal processing based on diversity combining� which will

be discussed in more detail below the phase�coherent methods have been tested in various

underwater environments The achieved data rates of up to � kbps over long range chan�

nels� and up to �� kbps over shallow water medium�range channels� are among the highest

reported to date

��� Design example� multisensor signal processing for coherent

detection

In many of the underwater acoustic channels multipath structure may exhibit one or more

components which carry the energy similar to that of the principal arrival As the time

progresses� it is not unusual for these components to exceed in energy the principal arrival

���� ��� The fact that the strongest multipath component may not be well dened makes

the extraction of carrier reference a di�cult task in the underwater channel To establish

coherent detection in the presence of strong multipath� a technique based on simultaneous

synchronization and multipath compensation has been presented in ���� This technique is

based on joint estimation of the carrier phase and the parameters of a decision�feedback

equalizer DFE�� where the optimization criterion is minimization of the mean�squared error

MSE� in the data estimation process In ����� the equalizer�synchronizer structure has been

extended to include a number of input array channels This implementation with spatial

diversity combining has shown superior performance in a variety of channels� as well as

��

potentials for dealing with several types of interference In Fig�� the multichannel equalizer

is shown� preceded by an additional pre�combiner� which may or may not be used depending

on the application and the number of available received channels

The input signals to the baseband processor are the A�D converted array signals� brought

to baseband using nominal carrier and lowpass ltering� and frame�synchronized using a

known channel probe usually a short sequence transmitted in phase and quadrature at the

data rate� Baseband processing begins with downsampling� which may be carried out to as

few as � samples per symbol interval Ts � T���� since the signals are shaped at the trans�

mitter using raised�cosine spectrum shaping which limits their bandwidth to less than ��T

The method is suitable for an all�digital implementation For applications where transmit�

ter and receiver are not moving� but only drifting with water� no explicit adjustment of the

sampling clock is needed It will implicitly be accomplished during the process of adaptive

fractionally spaced equalization The front section of the equalizer will also perform adaptive

matched ltering and linear equalization To correct for the carrier o�set� the signals in all

channels are phase�shifted by the amount estimated in the process of joint equalization and

synchronization After coherent combining� the ISI resulting from the previously transmit�

ted symbols postcursors� is cancelled in the feedback section of the equalizer This receiver

structure is applicable to any linear modulation format� such as M�PSK� or M�QAM� the

only di�erence being in the way in which symbol decision is performed In addition to com�

bining and equalization� signal processing at the receiver includes the operation of decoding

if the signal at the transmitter was encoded Trellis coded modulation� compatible with PSK

and QAM signals� provides an e�ective means of improving performance on a band�limited

channel ���� In addition to coded modulation� error correction coding may be employed

The receiver parameters that are adaptively adjusted are the weights of the pre�combiner�

the tap�weights of the feedforward lters� the carrier phase estimates� and the tap�weights of

the feedback lter A single estimation error is used for the adaptation of all the parameters

This error is the di�erence between the estimated data symbol which is input to the decision

device� and its true value During the initial training mode� the true data symbols are

known After the training period� when the receiver parameters have converged� the on�line

symbol decisions are fed back to the equalizer and used to compute the error The adaptive

��

algorithm which has proven to be e�ective in di�erent underwater acoustic environments

is a combination of the second�order digital phase�locked loop PLL� for the carrier phase

estimates� and the RLS algorithm for the multichannel equalizer tap weights The complexity

of the multichannel equalizer grows with the number of receiver array sensors For this

reason� the spatial pre�combiner may be used to limit the number of equalizer channels� but

still make use of the diversity gain The pre�combiner weights can be estimated jointly with

the rest of adjustable parameters The details of the joint adaptation are given in ����

The receiver is adaptively adjusted to coherently combine the multiple signal arrivals�

and thus exploit both the spatial and the temporal� or multipath diversity gain In this

manner� it di�ers from a receiver based on adaptive beamforming which is adjusted to

null out the signal replicas arriving from angles di�erent than that of the desired path

The signal isolated in this manner usually has to be processed by a separately optimized

equalizer� needed to compensate for the residual ISI which arizes because the beamformer

cannot completely eliminate the interference represented by the broad�band replicas of a

high�rate signal Since it is not constrained by angular resolution� the method of multichannel

equalization may be used with as few as two input channels� and is equally applicable to

a variety of underwater acoustic channels� regardless of the channel�s range�to�depth ratio

In applications where large arrays are available� the use of a pre�combiner reduces receiver

complexity� while preserving the multichannel diversity gain

The method of multichannel equalization was successfully used to process the signals

transmitted over a variety of underwater channels Experimental results include data rates

of � kbps over three convergence zones ��� nautical miles� in deep water� � kbps over ��

nautical miles in shallow water� and �� kbps over � nautical mile in a rapidly varying shallow

water channel

� Active research topics

At this stage in the development of underwater acoustic communication techniques� with

the feasibility of high rate communications established� a number of research topics are

foreseen which will in�uence the development of future systems These topics include

��

reduced�complexity receiver structures and algorithms suitable for real�time implementa�

tion� techniques for interference suppression� multiuser underwater communications� system

self�optimization� development of modulation�coding methods for improved bandwidth e��

ciency� and mobile underwater acoustic communication systems

�� Reducing the receiver complexity

Although the underwater acoustic channels are generally conned to low data rates as com�

pared to the radio channels�� the encountered channel distortions require complex receiver

structures� resulting in high computational load which may exceed the capabilities of the

available programmable DSP platforms Consequently� reducing the receiver complexity to

enable e�cient real�time implementation has been a focus of many recent studies

The problem of reducing the receiver complexity may be addressed on two levels� the

algorithmic and the structural level For application in time�varying channels� the receiver�

whether it is based on multichannel combining� equalization� or both methods� must use an

adaptive algorithm for adjusting its parameters Two commonly used types of algorithms

are based on the LMS and the RLS estimation principles

In a majority of recent studies� the LMS�based algorithms are considered an only al�

ternative due to their low computational complexity linear in the number of coe�cients

N� ���������� ���� However� the LMS algorithm has a convergence time which may become

unacceptably long when large adaptive lters are used �� N as opposed to � N of the RLS

algorithm� The total number of coe�cients N is usually large in shallow water medium

�and long�range applications more than ��� taps is often needed for spatial and temporal

processing� In addition� the standard LMS is very sensitive to the choice of the step�size

To overcome this problem� self�optimized LMS algorithms may be used ����� but this results

in increased complexity� and increased convergence time

RLS algorithms� on the other hand� have better convergence properties but higher com�

plexity The quadratic complexity of the standard RLS algorithm is too high when large

adaptive lters need to be implemented In general� it is desirable that the algorithm be

of linear complexity number of computations per iteration proportional to the number of

parameters�� a property shared by the fast RLS algorithms A numerically stable fast RLS

��

algorithm based on ���� has been used for o��line signal processing ���������� while the real�

time implementation ���� uses a square�root RLS algorithm of ���� Despite its quadratic

complexity� the advantage of this algorithm is that it allows the receiver parameters to be up�

dated only periodically� rather than every symbol interval� thus reducing the computational

load per each detected symbol In addition� the updating intervals can be determined adap�

tively� based on monitoring the mean squared error Such adaptation methods are especially

suited for use with high transmission rates At high symbol rates� the long ISI requires large

adaptive lters� therefore increasing the computational complexity However� the time inter�

val between two transmitted symbols decreases� eliminating the need to update the receiver

parameters every symbol interval Also� the square�root RLS algorithm has excellent numer�

ical stability� which makes it a preferable choice in many applications A di�erent class of

adaptive lters� which also have the desired convergence and numerical stability properties�

are the lattice lters which use RLS algorithms These algorithms have been proposed in

����� but have not yet been applied to underwater acoustic channel equalization Choosing

an appropriate receiver adaptation method will receive more attention in the future acoustic

modem design

Regardless of the adaptive algorithm used� its computational complexity is proportional

to the number of receiver parameters tap�weights� Rather than focusing on low�complexity

algorithms only� one may search for a way to reduce the receiver size Although the use of

spatial combining reduces residual ISI and allows shorter length equalizers to be used� a

broadband combiner may still require a large number of taps to be updated� limiting the

practical number of receiving channels to only a few In ���� a method is presented which

allows multichannel equalization of a large number of input channels This is accomplished

by using the multichannel equalizer of Fig� in which the spatial pre�combiner reduces a large

number of input channels to a smaller number for subsequent multichannel equalization The

parameters of the spatial pre�combiner and those of the multichannel equalizer are optimized

jointly to minimize the MSE in the data symbol estimate More than one channel at the

output of the combiner is usually required to achieve a substantial processing gain� however�

the reduced number of channels at which the full processing gain may be achieved is often

low eg� three�� leading to an interesting conclusion that reduction in complexity may be

��

achieved at no cost in performance This is explained by the fact that the multipath structure

is not independent among the array sensors In addition to the reduced computational

complexity� smaller adaptive lters result in less noise enhancement� contributing to improved

performance

A di�erent approach in the design of reduced�complexity receivers has been investigated

in ����� where the focus is made on reducing the number of equalizer taps A conventional

equalizer is designed to span all of the channel response However� if the channel is character�

ized by several distinct multipath arrivals separated in time by intervals of no reverberation�

an equalizer may be designed to have fewer taps� corresponding only to the signicant por�

tions of the channel response Such method� termed sparse equalization� was applied to

detection of QPSK signals recorded in the Arctic waters Results reported in ���� show an

order of magnitude reduction in computational load By reducing the number of adaptively

adjusted parameters� this approach also makes it possible to use simple updating algorithms�

such as standard RLS algorithms which have good numerical stability Finally� in channels

which are naturally sparse� discarding the low�magnitude equalizer taps in fact results in

improved performance since no unnecessary noise is processed

�� Interference cancellation and multiuser communications

The sources of interference in underwater acoustic channels include external interference and

internal interference that is generated within the system The external sources include noise

coming from on�board machinery or other nearby acoustic sources� as well as the propulsion

and �ow noise associated with the underwater vehicle launch process The internal noise�

which has signal�like characteristics� arises in the form of echo in full�duplex systems� or

multiple�access interference generated by other users operating within the same network

Recent advances in noise cancellation techniques for underwater acoustic communication

systems have been presented in ���� In this study the performance of several methods for

cancellation of band�limited white noise and multiple sinusoidal interference were investi�

gated It was found that the same receiver structure as that of Fig� was most e�ective in

cancelling the interference and noise while simultaneously detecting the desired signal Noise

cancellation is performed simply by feeding a reference of the noise signal to one of the mul�

��

tichannel combiner inputs� while cancellation of the sinusoidal interferer may be performed

even without the reference signal By virtue of having the training sequence� the multichan�

nel combiner has the capability to adaptively notch interfering signal out� and extract the

desired signal

A multiple�access communication system represents a special case of structured inter�

ference environment Due to the bandwidth limitation of the underwater acoustic channel�

frequency�division multiple�access may not be an e�cient technique Time�division multiple

access is associated with the problem of e�cient time�slot allocation which arises because of

the long propagation delays A possible solution in such a situation is to allow a number of

users to transmit simultaneously in both time and frequency The receiver then has to be

designed to deal with the resulting multiple�access interference� which may be very strong

in an underwater acoustic network The fact that transmission loss varies signicantly with

range� and that only very low code�division processing gains are available due to bandwidth

constraints� both contribute to the enhanced near�far e�ect in the underwater acoustic chan�

nel The multiuser detection methods suitable for underwater acoustic channels rely on

the principles of joint synchronization� channel equalization and multiple�access interference

cancellation ���� Two categories of receivers considered are centralized� in which the signals

of all the users are detected simultaneously eg� up�link reception at a surface buoy which

serves as a central network node�� and decentralized� in which only the desired user�s signal

needs to be detected eg� down�link reception by an ocean�bottom node� Similarly as in

the case of interference cancellation� the adaptive multichannel receiver of Fig� was shown

to have excellent capabilities in the role of a decentralized multiuser detector� operating

without any knowledge of the interfering signal Multiuser detection techniques have been

shown e�ective in tests with experimental data Array processing plays a crucial role in

the detection of multiuser signals� but is also associated with the problem of computational

complexity

�� System self�optimization

A receiver algorithm must use a number of parameters which need to be adjusted according

to the instantaneous channel conditions prior to turning the receiver on These parameters

��

include the number and location of array sensors which provide good signal quality� the sizes

of the equalizer lters� and their tracking parameters Their optimal value depends not only

on the general link conguration and location� but also on the time of operation In addition�

an increase the background noise level� caused for example by a passing ship� temporarily

disable the communication If the adaptive receiver algorithms are to be used in autonomous

systems� external assistance in algorithm parameter selection� receiver initialization� training

and synchronization should be minimized It is for this reason that self�optimized receiver

algorithms should be addressed in the future research

The rst steps in this direction are evident in the implementation of the self�optimized

LMS algorithms for an adaptive beamformer ���������� which eliminates external tuning of the

step�size for varying channel conditions� and the periodically updated RLS ����� self�adjusted

to keep a predetermined level of performance by increasing the tracking rate if the channel

condition worsens These strategies provide the receiver with the capability to adjust to the

ne channel changes However� they depend on the availability of a reliable reference of the

desired signal Since a training sequence is inserted only so often in the transmitted signal�

if the receiver looses convergence during detection of a data packet� the entire packet will

be lost An alternative to periodic re�insertion of known data� which increases the overhead�

methods for self�optimized or blind recovery may be considered

A blind equalization method based on using the cyclostationary properties of oversam�

pled received signals has recently been developed ����� which requires only the estimation of

second�order signal statistics to recover the data sequence in the absence of clock synchro�

nization Originally developed for linear equalizers� this method has been extended to the

case of decision�feedback equalizer� necessary for application in underwater acoustic channels

with extreme multipath These methods have proven successful in preliminary tests with

real data ��� Further work on blind system recovery for underwater acoustic channels will

focus on methods for array processing and carrier phase tracking

� Modulation and coding

Achieving high throughputs over band�limited underwater acoustic channels is conditioned

on the use of bandwidth�e�cient modulation and coding techniques ���� The results doc�

��

umented in contemporary literature are conned to signaling schemes whose bandwidth

e�ciency is at most � to � bps�Hz Higher level signal constellations� together with trellis

coding are being considered for use in underwater acoustic communications While trellis�

coded modulation is well suited for the vertical channels which have minimal dispersion� their

use on the horizontal channels requires further investigation In the rst place� conventional

signal mapping into a high�level PSK or QAM constellation may be associated with increased

sensitivity of detection on a time�varying channel Recent results in radio communications

show that certain types of high�level constellations are more robust to the channel fading and

phase variations than the conventional rectangular QAM constellations ���� The design of

codes to accompany these modulations is yet to be developed Another issue associated with

the use of coded modulation on the channels with long ISI is the receiver design which takes

full advantage of the available coding gain Namely� the delay in decoding poses problems

for an adaptive equalizer which relied on the feedback of instantaneous decisions Receiver

structures which deal with this problem as it applies to underwater channels are a subject

of current studies

In addition to bandwidth�e�cient modulation and coding techniques� the future under�

water communication systems will rely on data compression algorithms to achieve high data

rates over severely band�limited underwater acoustic channels This is another active area of

research� where current achievements ���� report on the development of algorithms capable of

achieving compression ratios in excess of one hundred E�cient data compression� together

with sophisticated modulation and coding techniques� is expected to make it possible for the

rst time to transmit underwater images in real time

� Mobile underwater communications

The problem of channel variability� already present in applications with a stationary trans�

mitter and receiver� becomes a major limitation for the mobile underwater acoustic commu�

nication system The ratio of the vehicle speed to the speed of sound ����� for a vehicle

speed of �� knots or �� km�h� many times exceeds its counterpart in the mobile radio chan�

nels ����� for a mobile moving at ��� km�h�� making the problem of time�synchronization

very di�cult in the underwater acoustic channel Apart from the carrier phase o�set� the

��

mobile underwater acoustic systems will have to deal with the motion�induced pulse com�

pression�dilation Latest results report on successful missions of experimental AUVs which

use commercial M�FSK acoustic modems for vehicle�to�vehicle communication ���� For

phase�coherent systems� algorithms for continuous tracking the time�varying symbol delay

in the presence of underwater multipath are currently under development

While many problems remain to be solved in the design of high�speed acoustic commu�

nication systems� recent advances in this area serve as an encouragement for future work�

which will enable the capability to remotely explore the underwater world

��

References

��� LBerkhovskikh and YLysanov� Fundamentals of Ocean Acoustics� New York�

Springer� ����

��� SFlatte� Ed� Sound Transmission Through a Fluctuating Ocean� Cambridge� UK�

Cambridge University Press� ����

��� AQuazi and WKonrad� �Underwater acoustic communications�� IEEE Comm�

Magazine� pp ������ Mar ����

��� JCatipovic� �Performance limitations in underwater acoustic telemetry�� IEEE J�

Oceanic Eng�� vol ��� pp �������� July ����

��� ABaggeroer� �Acoustic telemetry � an overview��IEEE J� Oceanic Eng�� vol �� pp

�������� Oct ����

��� MStojanovic �Recent advances in high rate underwater acoustic communications��

IEEE J� Oceanic Eng�� pp�������� Apr ����

��� IEEE J� Oceanic Eng�� Special Issue on Image Processing for Oceanic Applications�

vol��� Jan ����

��� RHOwen� BVSmith and RFWCoates� �An experimental study of rough surface

scattering and its e�ects on communication coherence�� in Proc OCEANS���� pp

III����III���� Brest� France� Sept ����

��� AEssebbar� GLoubet and FVial� �Underwater Acoustic Channel Simulations for

Communication�� in Proc OCEANS���� pp III����III���� Brest� France� Sept

����

���� AFalahati� BWoodward and SBateman� �Underwater acoustic channel models for

���� b�s QPSK signals�� IEEE J� Oceanic Eng�� vol��� pp������ Jan����

���� CBjerrum�Niese� LBjorno� MAPinto and BQuellec� �A simulation tool for high

data�rate acoustic communication in a shallow�water� time�varying channel�� IEEE

J� Oceanic Eng�� vol ��� pp�������� Apr ����

��

���� JCatipovic� MDe�enbaugh� LFreitag and DFrye� �An acoustic telemetry system

for deep ocean mooring data acquisition and control�� in Proc� OCEANS���� pp

�������� Seattle� Washington� Oct ����

���� SCoatelan and AGlavieux� �Design and test of a multicarrier transmission system

on the shallow water acoustic channel�� in Proc OCEANS���� pp III����III����

Brest� France� Sept ����

���� AKaya and SYauchi� �An acoustic communication system for subsea robot�� in

Proc� OCEANS���� pp �������� Seattle� Washington� Oct ����

���� MSuzuki and TSasaki� �Digital acoustic image transmission system for deep sea

research submersible�� in Proc� OCEANS���� pp �������� Newport� RI� Oct ����

���� DHoag and VKIngle� �Underwater video compression using the wavelet trans�

form�� in Proc OCEANS���� San Diego� CA� Oct ����

���� AGoalic et al� �Toward a digital acoustic underwater phone�� in Proc OCEANS����

pp III����III���� Brest� France� Sept ����

���� BWoodward and HSari� �Digital underwater voice communications�� IEEE J�

Oceanic Eng�� vol ��� pp�������� Apr ����

���� JFischer et al� � A high rate� underwater acoustic data communications

transceiver�� in Proc� OCEANS���� pp �������� Newport� RI� Oct ����

���� RFWCoates� MZheng and LWang� �BASS ��� PARACOM� A �model� underwa�

ter parameteric communication system�� IEEE J� Oceanic Eng�� vol ��� pp��������

Apr ����

���� GSHowe et al� �Sub�sea remote communications utilising an adaptive receiving

beamformer for multipath suppression�� in ProcOCEANS���� pp I����I���� Brest�

France� Sept ����

��

���� MStojanovic� JACatipovic and JGProakis� �Phase coherent digital communica�

tions for underwater acoustic channels�� IEEE J� Oceanic Eng�� vol ��� pp ��������

Jan ����

���� MStojanovic� JACatipovic and JGProakis� �Adaptive multichannel combining

and equalization for underwater acoustic communications �� Journal of the Acoustical

Society of America� vol �� ��� Pt �� pp ���������� Sept ����

���� MStojanovic� JACatipovic and JGProakis� �Reduced�complexity multichannel

processing of underwater acoustic communication signals�� Journal of the Acoustical

Society of America� vol �� ��� Pt �� pp�������� Aug ����

���� MStojanovic� JGProakis and JA Catipovic� �Performance of a high rate adaptive

equalizer on a shallow water acoustic channel�� J� Acoust� Soc� Amer�� vol ��� ���

Pt �� pp ���������� Oct ����

���� GAyela� MNicot and XLurton� �New innovative multimodulation acoustic commu�

nication system�� in Proc OCEANS���� pp I����I���� Brest� France� Sept ����

���� MJohnson� DHerold and JCatipovic� �The design and performance of a compact

underwater acoustic network node�� in Proc OCEANS���� pp III�������� Brest�

France� Sept ��

���� DSlock and TKailath� �Numerically stable fast transversal lters for recursive least

squares adaptive ltering�� IEEE Trans� Sig� Proc�� vol SP���� pp ������� Jan

����

���� FHsu� �Square root Kalman ltering for high�speed data received over fading dis�

persive HF channels�� IEEE Trans� Inform� Theory� Vol IT���� pp �������� Sept

����

���� BGeller� VCapellano� J�MBrossier� AEssebar and GJourdain� �Equalizer for

video rate transmission in multipath underwater communications�� IEEE J�Oceanic

Eng�� vol ��� pp�������� Apr ����

��

���� FLing and JGProakis� �Adaptive lattice decision�feedback equalizers�their perfor�

mance and application to time�variant multipath channels�� IEEE Trans� Commun��

vol ��� pp �������� Apr ����

���� MKocic� DPBrady and MStojanovic� �Sparse equalization for real�time digital

underwater acoustic communications�� in Proc OCEANS���� San Diego� CA� Oct

����

���� JCatipovic� MJohnson and DAdams� �Noise cancelling performance of an adap�

tive receiver for underwater communications�� in Proc ���� Symposium on AUV

Technology� pp �������� Cambridge� MA� July ����

���� MStojanovic and ZZvonar� �Multichannel processing of broadband multiuser com�

munication signals in shallow water acoustic channels�� IEEE J� Oceanic Eng�� pp

�������� Apr ����

���� LTong� GXu and TKailath� �Blind identication and equalization based on second�

order statistics��IEEE Trans� Inform� Theory� vol IT���� pp �������� Mar ����

���� JProakis� �Coded modulation for digital communications over Rayleigh fading chan�

nels�� IEEE J� Oceanic Eng� vol ��� pp ������ Jan ����

���� WTWebb and RSteele� �Variable rate QAM for mobile radio�� IEEE Trans� Com

mun�� Vol COM���� pp ���������� July ����

���� SChappell et al� �Acoustic communication between two autonomous underwater

vehicles�� in Proc ���� Symposium on AUV Technology � pp �������� Cambridge�

MA� ����

��

0 2 4 6 8 10 12 14 16 18 200

10

20

30

40

50

60

70

5km

10km

50km

100km

frequency [kHz]

rela

tive

SN

R [d

B]

Figure �� Frequency�dependent portion of SNR

0 5 10 15 200

5

10

15

20

25

wind speed [m/s]

Dop

pler

spr

ead

[Hz]

20 kHz

10 kHz

5 kHz

1 kHz

Figure �� Doppler spread on a single surface�re�ected path

��

0

5

10

15

-40

-20

0

20

40

0

0.2

0.4

0.6

0.8

time [s]delay [ms]

tx depth : 100 m, rx depth : 640 mChannel # 6 : omnidirectionalRate : 333 spsRange : 110 nautical miles

Figure �� Ensemble of long�range channel responses in deep water approx ���� m� o� the

coast of California� during the month of January Carrier frequency is � kHz

0

5

10

15-50

0

50

0

0.2

0.4

0.6

0.8

time [s]delay [ms]

tx depth : 25 m, rx depth : 23 mChannel # 8 : omnidirectionalRate : 333 spsRange : 48 nautical miles

Figure �� Ensemble of long�range channel responses in shallow water approx �� m� o� the

coast of New England� during the month of May Carrier frequency is � kHz

��

time [s]delay [ms]

tx depth : 8 m, rx depth : 3.5 m

Channel # 1 : directional

Rate : 500 sps

Range : 2 nautical miles

02

46

810

-10

-5

0

5

100

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Figure �� Ensemble of medium�range channel responses in shallow water approx �� m� o�

the coast of New England� during the month of February Carrier frequency is �� kHz

input K���

�resmpl�clockK

m��

�

carriercorrect�

K

input ����

Ns�T

�resmpl�clock�

m��

�

carriercorrect�

� spatialpre�

combiner

�

�

feed�forwardlter K

feed�forwardlter �

�

�

��

��

��T

�

�

��� � m� � symbol

decision�

data out

�feedbacklter

��

� trellisdecoder

data out

�with coding �

Figure �� A multichannel receiver for phase�coherent detection

��

Contents

�� System requirements � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

� Channel characteristics �

�� Range� bandwidth and SNR � � � � � � � � � � � � � � � � � � � � � � � � � � � �

�� Multipath � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

�� Time�variation � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �

� System design �

�� Systems based on noncoherent modulation � � � � � � � � � � � � � � � � � � � ��

�� Systems based on coherent modulation � � � � � � � � � � � � � � � � � � � � � ��

� Signal processing methods for multipath compensation ��

�� Design example� multisensor signal processing for coherent detection � � � � ��

� Active research topics ��

�� Reducing the receiver complexity � � � � � � � � � � � � � � � � � � � � � � � � ��

�� Interference cancellation and multiuser communications � � � � � � � � � � � � ��

�� System self�optimization � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ��

�� Modulation and coding � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ��

�� Mobile underwater communications � � � � � � � � � � � � � � � � � � � � � � � ��

List of Figures

� Frequency�dependent portion of SNR � � � � � � � � � � � � � � � � � � � � � � ��

� Doppler spread on a single surface�re�ected path � � � � � � � � � � � � � � � ��

� Ensemble of long�range channel responses in deep water approximately ����

m� o� the coast of California� during the month of January Carrier frequency

is � kHz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ��

��

� Ensemble of long�range channel responses in shallow water approximately ��

m� o� the coast of New England� during the month of May Carrier frequency

is � kHz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ��

� Ensemble of medium�range channel responses in shallow water approximately

�� m� o� the coast of New England� during the month of February Carrier

frequency is �� kHz � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � ��

� A multichannel receiver for phase�coherent detection � � � � � � � � � � � � � ��

��